National Center for Emerging and Zoonotic Infectious Diseases

NHSN Analysis: Focus on Device-Associated Data

Zuleika Aponte-Torres, MPH, MCPH

March 21, 2017
Learning Objectives
• Review the standardized infection ratio (SIR) and its use in the
interpretation of Device-associated data.

• Describe the risk-adjustment methods for Device-associated Data.

• Help you understand your data and how to analyze it.

Why Analyze?
 Analysis tools within NHSN help facilitate internal validation activities and
help ensure accuracy!

 Reports generated from NHSN can help inform prioritization and success
of prevention activities.

 Data entered into NHSN may be used by: CDC, CMS, your state health
department, your corporation, special study groups, etc.

 These are YOUR data!!!!!

A Review: The Standardized Infection Ratio (SIR)
 SIR – A summary statistic that compares the number of healthcare-
associated infections (HAIs) that were reported to the number of HAIs that
were predicted to occur, based on a calculation using data for HAI events
that occurred in a given referent time period.

# 𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐𝒐 𝑯𝑯𝑯𝑯𝑯𝑯𝒐𝒐
SIR =
# 𝒑𝒑𝒐𝒐𝒐𝒐𝒐𝒐𝒑𝒑𝒑𝒑𝒑𝒑𝒐𝒐𝒐𝒐 𝑯𝑯𝑯𝑯𝑯𝑯𝒐𝒐

*The baseline I'll be discussing today is the new 2015 baseline and risk adjustment.
A Review: The Standardized Infection Ratio (SIR) and
National SIR Baseline
 SIR interpretation:
– 1 = same number of infections reported as would be predicted given
the US baseline data
– Greater than 1= more infections reported than what would be
predicted given the US baseline data
• SIR of 1.25 = 25% more infections than predicted
– Less than 1 = fewer infections reported than what would be predicted
given the US baseline data
• SIR of 0.50 = 50% fewer infections than predicted
Basis for Using SIRs and not Rates
 The SIR allows users to summarize data by more than a single stratum (e.g.
location or procedure category), adjusting for differences in the incidence
of infection among the strata.

 The SIR permits comparisons between the number of infections

experienced by a facility, group, or state to the number of infections that
were predicted to have occurred based on national data (i.e., baseline
data).
How many of you have read the SIR Guide?
Number of Predicted Infections-Device
Associated
 The pooled mean is used to calculate the number
of predicted events for CLABSI?
A. No
B. Yes

Previous predicted # calculation: New 2015 baseline predicted # calculation:

• For CLABSI and CAUTI SIRs, the predicted # is
calculated for each individual location as:
# device days *(NHSN pooled mean/1000)
• Where the pooled mean originates from a
defined baseline report.
Calculating the Number of Predicted Infections
 The number of predicted infections in NHSN is calculated based on the
2015 national HAI aggregate data and adjusted for each facility using
variables found to be significant predictors of HAI incidence.
 Negative binomial regression models are used to calculate the number of
predicted events for CLABSI, MBI-LCBI, CAUTI, VAE.

https://www.cdc.gov/nhsn/pdfs/ps-analysis-resources/nhsn-sir-guide.pdf
Using Models for Device-associated Infections
 General Negative Binomial Regression Model:

A Guide to the SIR: https://www.cdc.gov/nhsn/pdfs/ps-analysis-resources/nhsn-sir-guide.pdf

Factors Included in the Model: Acute Care
Hospitals (ACHs)
Factors Included in the Model: Acute Care Hospitals
(ACHs) (non-NICU)
Factor CLABSI CLABSI CAUTI Total VAE IVAC
(NICU) Plus
CDC Location    

Facility Type   

Medical School Affiliation*   

Birthweight 

Facility Bed size*   

* Variables taken from the Annual Survey

 Methods for Calculating the Predicted Number of
CLABSI Infections in Acute Care Hospital (non-NICU)
Effect Parameter Standard Error p-value
Estimate
Intercept -7.339 0.0190 <0.0001
CDC Location: Burn Critical Care 1.2874 0.1119 <0.0001
CDC Location: Medical Critical Care 0.2539 0.0405 <0.0001
Location: Other Adult ICUs and Mixed Acuity 0.1164 0.0197 <0.0001
CDC Location: Pediatric Cardiothoracic Critical Care 0.4130 0.0595 <0.0001
CDC Location: Trauma Critical Care 0.5411 0.0744 <0.0001
Location: Adult Oncology Units (ICU and Ward) 0.2755 0.0367 <0.0001
Location: Pediatric Oncology Units ICU and Ward) 0.1960 0.0867 0.0237
Location: Adult and Pediatric Wards (ref) REFERENT - -
Facility Bed Size* ≥296 0.1686 0.0195 <0.0001
Facility Bed Size* <296 REFERENT - -
Medical School Affiliation*: Major, Graduate, or Undergraduate
0.1958 0.0197 <0.0001
Teaching Status
Medical School Affiliation*: Non-Teaching Hospital REFERENT - -
Facility Type: Children's/Women's Hospital 0.1328 0.0643 0.0388
Facility Type: All Other REFERENT - -
* Variables taken from the Annual Survey
Example: Applying Risk Model for CLABSI, NHSN 2015
= exp(-7.339 +
1.2874*(Burn Critical Care* ) +
0.2539*(Medical Critical Care* ) +
0.1164*(Other Adult ICUs and Mixed Acuity*) +
0.413*(Pediatric Cardiothoracic Critical Care* ) +
0.5411*(Trauma Critical Care* ) +
0.2755*(Adult Oncology Units (ICU and Ward*) +
0.196*(Pediatric Oncology Units (ICU and Ward*) +
0.1686* (Facility Bed Size: ≥ 296 beds*) +
0.1958* (Medical School Affiliation: Major,
Graduate, or Undergraduate Teaching Status*)+
0.1328*(Facility Type: Children's/Women's
Hospital Children’s Hospital, Women’s Hospital,
Women’s/Children’s Hospital*)) * numCLDays

* For these risk factors, if present = 1; if not = 0

Example: Applying Risk Model for CLABSI, NHSN 2015
 Facility Profile: Monthly (February 2015)
– 135-beds
– General Acute Care Hospital
– Major teaching affiliation
– Reporting for:
• Mixed Age Mixed Acuity Unit
• Neurologic Critical Care
• Medical/Surgical Critical Care
– With 238 central line days and 1 CLABSI event reported for the month
of February
Example: Applying Risk Model for CLABSI, NHSN 2015
New model for calculation the number of
predicted CLABSI events:
= exp(-7.339 +
1.2874*(0) +
0.2539*(0) +
0.1164*(1) +
0.413*(0) +
0.5411*(0) +
0.2755*(0) +
0.196*(0) +
0.1686* (0) +
0.1958* (1)+
0.1328*(0)) * 238
= 0.211 predicted CLABSI events for the
month of February
Example: Applying Risk Model for CLABSI, NHSN 2015
New model for calculation the number of
predicted CLABSI events:
= exp(-7.339 +
1.2874*(0) +
0.2539*(0) +
0.1164*(1) +
0.413*(0) +
0.5411*(0) +
0.2755*(0) +
0.196*(0) +
0.1686* (0) +
0.1958* (1)+
0.1328*(0)) * 238
= 0.211 predicted CLABSI events for the
month of February * For these risk factors, if present = 1; if not = 0
 Your hospital has been measuring CLABSI using SIRs
under the previous, 2006-2008 baseline. Should you
compare these SIRs with those calculated under the
2015 baseline?
A. No
B. Yes
• SIRs under the original baseline cannot be directly compared to any SIRs calculated under the new
baseline, because of the different risk adjustment, different baseline population.
• When comparing SIRs from two time periods, both SIRs must have been calculated under the same
baseline, for example:
• 2014 vs. 2015 SIRs: original baseline
• 2015 vs. 2016 SIRs: use either the new baseline or original baseline for both SIRs in the
comparison
• 2016 vs. 2017 SIRs: new baseline
Standardized Infection Ratio (SIR) Table

 The following example:

– shows how to calculate and interpret the SIR for device-associated
infections
– is for CLABSI, but the CAUTI SIR is calculated and interpreted in a
similar manner
Standardized Infection Ratio (SIR) Table
Standardized Infection Ratio (SIR) Table
 Report Modification: For the purpose of this example, the modifications that have been
made are: summaryYr was set to 2015, filtered by Critical Care (CC) locationType, and the
report grouped by summaryYH.
Standardized Infection Ratio (SIR) Table
 Report Modification: For the purpose of this example, the modifications that have been
made are: summaryYr was set to 2015, filtered by Critical Care (CC) locationType, and
the report grouped by summaryYH.
Standardized Infection Ratio (SIR) Table
 Report Modification: For the purpose of this example, the modifications that have been
made are: summaryYr was set to 2015, filtered by Critical Care (CC) locationType, and the
report grouped by summaryYH.
Standardized Infection Ratio (SIR) Table
 Output/Results
Standardized Infection Ratio (SIR) Table-Interpretation

 This facility reported 4 central line-associated BSI (infCount) in critical care locations
(locationType=”CC”) during the first half of 2015. This is the observed number of CLABSIs.

 The overall SIR for this facility during this time period is 2.660, indicating that this facility
observed approximately 166% more infections than predicted. The number of CLABSIs
predicted to occur for the first half of 2015 is 1.504 and 0.011 for the second half.

 An SIR will only be calculated if the number of predicted infections is ≥ 1.

Standardized Infection Ratio (SIR) Table

Interpretation:
 In this example, the p-value for the first half of 2015 is greater than 0.05 and thus there is no
significant difference between the number of infections observed and the number of
infections predicted.
 If the confidence interval includes the value of 1, then the SIR is not significant (the number
of observed infections is not significantly different from the number predicted, using the
same convenient cut point).
 The statistical evidence should be interpreted as insufficient to conclude that the SIR is
different than 1.
The Re-baseline: Will my SIRs change?
 In short…Yes.
 In addition to different risk models being used, the rebaselined SIRs will be
using data with different incidence than the first baseline.
The Re-baseline: Will my hospital’s SIRs change?
Annual, Facility-level CAUTI data from a Critical Access Hospital with an
undergraduate medical school affiliation
Baseline 1 (2009 NHSN Data):

Baseline 2 (2015 NHSN Data):

Fictitious data. Example provided for illustrative purposes only.

The Re-baseline: What to Expect with New Models
 CLABSI:
– Separate models for each setting (e.g., ACHs, LTACHs)
– MBIs will be excluded from the 2015 baseline
– Risk models will be used to assess predicted # of infections and will
incorporate a level of categorization by CDC location (for ACHs), and
relevant, significant facility-level factors
– Previously-excluded inpatient locations (e.g. Telemetry Ward, Mixed
Acuity Ward) will be included under the 2015 baseline.
Real World Case
 I need your assistance with a missing CLABSI for 2016 since the recent
upgrade.
– The CLABSI was captured on reports I ran prior to the upgrade.
– Missing is 1 CLABSI for my Hematology/Oncology Ward that appears
on our line list but not on the SIR reports.

Can you tell me why I’m missing this event?

Real World Case-Answer
 The reason why this CLABSI is being excluded from the calculations is
because this is MBI-LCBI event.
 With the re-baseline, MBI-LCBI events have been removed from the
CLABSI numerator.
 Later on this year, we will be introducing MBI-LCBI rate and SIR tables in
the NHSN application.
Checking for MBI-LCBI
 Run the CLABSI Frequency Table and include the variable mbi_lcbi
 Based on our results, 1 CLABSI would be excluded when our data will be
calculated using the 2015 baseline:

https://www.cdc.gov/nhsn/PS-Analysis-resources/PDF/MBIAnalysis.pdf
The Re-baseline: What to Expect with New Models
 CAUTI:
– Separate models for each setting (e.g., ACHs, LTACHs).
– Urinary catheter days will continue to be used in the SIR calculation
– Risk models will be used to assess predicted # of infections and will
incorporate a level of categorization by CDC location (for ACHs), and
relevant, significant facility-level factors
– Previously-excluded inpatient locations (e.g. Telemetry Ward, Mixed
Acuity Ward) will be included under the 2015 baseline.
CAUTI (and CLABSI) SIR Reports in NHSN
CAUTI (and CLABSI) SIR Reports in NHSN
 SIR Outputs will include 5 tables:
1. Overall SIR for the facility
2. SIR by location type
3. SIR by CDC location
4. SIR by individual locations
5. Data Not Included in the SIR
CAUTI (and CLABSI) SIR Output- Table 1
 This table will include all the units for which your hospital reported data
during that time period.
 For this example we use the cumulative group by function
CAUTI (and CLABSI) SIR Output- Table 2
 This table produces an SIR for each Location Type (eg. ICUs, WARDs)
CAUTI (and CLABSI) SIR Output- Table 3
 This table produces an SIR for each CDC location type that has CAUTI data
entered in the facility.
CAUTI (and CLABSI) SIR Output- Table 4
 This table produces an SIR for each individual location that has CAUTI data
entered in the facility.
CAUTI (and CLABSI) SIR Output- Table 5
 This table produces a list of the locations that are not included in the SIR
(eg. missing data or outpatient locations)
The Re-baseline: What to Expect with New Models
 VAE:
– Separate models for each setting (e.g., ACHs, LTACHs)
– Will be calculated for “Total VAE” as well as “IVAC Plus”
– Risk models will be used to assess predicted # of events and will
incorporate a level of categorization by CDC location (for ACHs), and
relevant, significant facility-level factors
– There are no VAE models and thus, no SIRs for: IRF (no Total VAE or
IVAC Plus model), CAH (no IVAC Plus model)
VAE SIR-Outputs
 Separate models for each setting (e.g., ACHs, LTACHs)
VAE SIR-Outputs
 Will be calculated for “Total VAE”
VAE SIR-Outputs
 As well as “IVAC Plus”
 If your facilit y had a Predicted Number of
infections equal to 0.896 for 2015Q2, will the
SIR be calculated for that quarter?
A. Yes
B. No
C. Maybe, it will depend on the number of events for that quarter

• When the predicted number of infections is <1, it is considered too low to calculate a precise
SIR and comparative statistics.
• When this occurs, you may wish to group your SIRs by a longer time period, such as calendar
year (summaryYr).
• Run the TAP Reports to review the CAD (cumulative attributable difference, which is the
difference between the # observed and # predicted).
Device Utilization Ratios
 Device utilization (DU) ratios help assess the proportion of days in which
patients were at risk for the DA infection

 Calculated as:

# 𝒐𝒐𝒇𝒇 𝒐𝒐𝒐𝒐𝒐𝒐𝒑𝒑𝒑𝒑𝒐𝒐 𝒐𝒐𝒂𝒂𝒚𝒚𝒐𝒐

# 𝒐𝒐𝒇𝒇 𝒑𝒑𝒂𝒂𝒑𝒑𝒑𝒑𝒐𝒐𝒏𝒏𝒑𝒑 𝒐𝒐𝒂𝒂𝒚𝒚𝒐𝒐
 Pooled Means (National Benchmark Rates)
 2014 was the last year NHSN published device-associated national pooled means
– Infection rate and device utilization ratio (DUR)
– Moving forward, benchmarks will be published annually as SIRs

 Typically, rate tables provided the facility’s rate

and DUR, with a comparison to national
pooled means
Rate Tables
 Pooled means will no longer appear in the default device-associated rate
tables for 2015 data and forward
 Rate Tables in “Baseline Set 1” Folder
 Review the comparison between your facility’s rates and the 2014 national pooled mean rates in the
“Baseline Set 1” folder
– 2015 + 2016 device-associated rates will be compared to 2014 national pooled mean

 Reminder: 2014 is the last pooled mean in the “Baseline Set 1” rate tables
 2015 national pooled means will be available in the Rate Calculator
Rate Calculator
 *New* online tool launching this year
 Public website outside of the NHSN application
 User will enter risk factors as they apply to the facility/HAI of interest
– e.g., bed size, medical school affiliation
 Calculator will produce a national pooled mean rate for the facility based on 2015
national data
– No annual updates
 All HAI types (including SSI, MRSA & C.difficile LabID, etc.)
Rate Calculator- Preview
 Exercise - CLABSI
 Your administration has asked you to provide a summary statistic
describing the CLABSI experience in your Acute Care Hospital for the first
half of 2015.
 You will need to be able to interpret the statistic and its associated tests of
statistical significance for the administrators.
 To be able to answer this question you will need to run the “SIR – Acute
Care Hospital CLAB Data” report.
 Exercise - CLABSI
 The SIR for the first half of 2015 is 2.326
 How can this overall CLABSI SIR of 2.326 be
interpreted?

A. This facility observed more infections than predicted.

B. This facility observed the same number of infections predicted.
C. This facility observed less infections than predicted.
D. None of the above

The overall SIR for this facility during this time period is 2.326, indicating that
this facility observed approximately 133% more infections than predicted.
What changes can potentially impact my SIRs?

 Entry or deletion of events

 Changes to number of patient days, device days, admissions
 Removal or addition to monthly reporting plans
 Changes to relevant factors in the annual survey (e.g.,medical school
affiliation, facility bedsize)
 Resolution of “Report No Events” alerts
Take-Home Points

 Updated risk adjustment will be applied across various HAI types and healthcare
settings
 SIRs produced under the new 2015 baseline will not be comparable to SIRs
calculated under the original baselines
 The 2015 baseline is a new “starting/referent point” from which to measure future
progress –therefore, we expect that hospital SIRs will shift closer to 1, particularly
for the 2015 SIRs calculated with the 2015 baseline
Device Associated Analysis Resources
 A Guide to the SIR: https://www.cdc.gov/nhsn/pdfs/ps-analysis-
resources/nhsn-sir-guide.pdf
 Analysis Output Quick Reference Guides: http://www.cdc.gov/nhsn/ps-
analysis-resources/reference-guides.html
 Analyzing MBI-LCBI CLABSI Data: https://www.cdc.gov/nhsn/PS-Analysis-
resources/PDF/MBIAnalysis.pdf
 Hospital Compare Data Verification: https://www.cdc.gov/nhsn/PS-
Analysis-resources/PDF/MBIAnalysis.pdf

Help with any analysis outputs: email [email protected]

 Thank You!
[email protected]

For more information, contact CDC

1-800-CDC-INFO (232-4636)
TTY: 1-888-232-6348 www.cdc.gov

The findings and conclusions in this report are those of the authors and do not necessarily represent the
official position of the Centers for Disease Control and Prevention.